TIGER-Lab/VLM2Vec-Full
Partial fine-tune vs microsoft/Phi-3.5-vision-instruct. Partial change: 90% of the weights changed (attention, embeddings, mlp), in 32 of 32 layers.
Tensors changed: Embeddings 92/398, Norms 0/65, MLP 64/64, Attention 64/64, lm_head 0/1